{"id":"W4407174976","doi":"10.3389/frai.2025.1533508","title":"Large language models generating synthetic clinical datasets: a feasibility and comparative analysis with real-world perioperative data","year":2025,"lang":"en","type":"article","venue":"Frontiers in Artificial Intelligence","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":24,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Calgary","funders":"","keywords":"Computer science; Fidelity; Data mining; Sample (material); Missing data; Sample size determination; Artificial intelligence; Machine learning; Statistics; Mathematics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01346076,0.00083537,0.0004827162,0.000881694,0.0003392468,0.001215072,0.001720104,0.000970934,0.001841932],"category_scores_gemma":[0.05390611,0.0002562831,0.001014824,0.0007182827,0.0007381274,0.001237395,0.001468479,0.0012387,0.0006659082],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008956865,"about_ca_system_score_gemma":0.001288503,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003578418,"about_ca_topic_score_gemma":0.003184776,"domain_scores_codex":[0.9935236,0.004502304,0.0004105249,0.0007437858,0.000669494,0.0001502213],"domain_scores_gemma":[0.9511212,0.04146304,0.0009103976,0.003581922,0.002399048,0.0005244242],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.004608816,0.002627884,0.07828546,0.00210207,0.0007768645,0.001281366,0.001890457,0.5970387,0.008576785,0.009047597,0.02072133,0.2730426],"study_design_scores_gemma":[0.0003027299,0.00108857,0.008446052,0.000127984,0.00008105712,0.0003263572,0.0006357504,0.973805,0.005587577,0.004782188,0.004758094,0.00005863244],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.818159,0.0009930463,0.1577203,0.002262838,0.0003787352,0.001284558,0.01179921,0.003921191,0.00348114],"genre_scores_gemma":[0.8659266,0.0003287199,0.1143465,0.000356518,0.00007918764,0.0007610904,0.01737843,0.0002069881,0.0006160968],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01346076,"threshold_uncertainty_score":0.07118815,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3388162493346323,"score_gpt":0.5062546044831657,"score_spread":0.1674383551485334,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}